CheckInGatedChanges Location of activity in the assembly workflow

Question

The default build template has the CheckInGatedChanges ("Check In Gated Changes") action, which runs outside the AgentScope ("Run on Agent") action.

Is there a reason for this? Rather, can the CheckInGatedChanges operation be transferred to the end of the AgentScope (end of sequence)?

I assume that Microsoft does not want the agent itself to register the changes, but the controller. I am not sure though, since I do not know the internal operation of the controller-agent model that they have.

The reason I'm asking is because execution requires execution after the changes, and this currently forces me to use a different instance of Agent Activity. I would rather just put them all in the first AgentScope activity and not use additional resources.

Answer 1

The current location of CheckInGatedChanges in the template ensures that if you add additional actions to the template, they will not start unexpectedly after the changes are fixed on the server. In addition, it allows parallel execution of the agent. For example, if you want to parallelize an assembly to multiple agents for scaling, validation should occur after all assembly agent worker processes have completed.

This is what I can think of. I think it’s safe if you know that you don’t need a converged point to check your changes, to move CheckInGatedChanges activity to AgentScope so that you do not have to run another AgentScope to do your individual work. I do not think that we limit the scope of this activity.